• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

(Anti) SFF fun house

There's the problem..

You stopped using the Asus :D

I just flashed to the latest bios for my board, 4602 and it seems pretty solid on both my X3D and 5900X..
 
There's the problem..

You stopped using the Asus :D

I just flashed to the latest bios for my board, 4602 and it seems pretty solid on both my X3D and 5900X..

I have been avoiding the Strix because I think there's something messed up with it; it couldn't run even my replacement 5700G properly, which is noticeably stronger than my old one.

Dunno, my head is just spinning at this point, the 30W issue seems to have no rhyme or reasoning whatsoever and just comes and goes whenever it pleases. I can't reproduce it when I want to, and when I think I've fixed it BAM it's back.

The only two other people I can find with the same issue are all on AM4, one with a 5500 and one 5600X. I think maybe it's time to get away from this toxic fucking relationship
 
I think maybe it's time to get away from this toxic fucking relationship
I still have my Strix B660i :D

I want to try a 14th Gen on it, probably nothing too crazy..
 
When I was new and knew nothing of AM4, I took the advice of you and a couple of others.. you should have just copied my build :D
 
I still have my Strix B660i :D

I want to try a 14th Gen on it, probably nothing too crazy..
A 13700k with a 76w PL1/PL2 is still a pretty potent option.

Definitely not in this form factor lmao........but I am considering it haha, maybe a cheap i3 or i5

Gonna slip in a board swap probably today. It's a major pain in the ass regardless of what I do, the HT5 is setup in APU mode so I have to swap the entire back panel of the case. Also not sure if Big Shuriken 3 clears the Strix-I's stupid IO shroud. And swapping the two HDPlex DC-ATX units are always a pain. And I have to remove the internal HDPlex AC-DC brick in the HT5......

I have no confidence in the board swap fixing anything at all. If it doesn't I might sell the 5700G and board and really consider Intel. Or just pack it in. I'm not really missing functionality anymore now with my G14. 680M kicks the absolute shit out of Vega 8.

@freeagent what CPU did you have?
 
Last edited:
Definitely not in this form factor lmao........but I am considering it haha, maybe a cheap i3 or i5

Gonna slip in a board swap probably today. It's a major pain in the ass regardless of what I do, the HT5 is setup in APU mode so I have to swap the entire back panel of the case. Also not sure if Big Shuriken 3 clears the Strix-I's stupid IO shroud. And swapping the two HDPlex DC-ATX units are always a pain. And I have to remove the internal HDPlex AC-DC brick in the HT5......

I have no confidence in the board swap fixing anything at all. If it doesn't I might sell the 5700G and board and really consider Intel. Or just pack it in. I'm not really missing functionality anymore now with my G14.
I have a spare cooler if you need :)

:D
 
I have a spare cooler if you need :)

:D

Nah I'm good, I just prefer to use the BS3, it's the best small cooler I have......my other option L9x65 is a bit of a screamer :laugh:
 
There's the problem..

You stopped using the Asus :D

I just flashed to the latest bios for my board, 4602 and it seems pretty solid on both my X3D and 5900X..

It's still early and I'm putting the Strix through its paces again, but so far it looks like Gigabyte fucked up royally with its boards......for whatever reason incapable/unwilling to supply enough slot power. Dunno what that's about, all of the complaints I can find about this specific issue are on Gigabyte AM4.

Sadly I have to use my L9x65 but for a different reason, I totally forgot about the different socket placement and the location of the barrel connector...
 
I totally forgot about the different socket placement and the location of the barrel connector...
All 3 of my Asus AM4 boards have slightly different socket placement too. My Strix -F actually has a lower Z height than the other two boards crazily enough..

I can tell with my Le Grand Macho RT.. nice and tight on 2 out of 3 boards..
 
All 3 of my Asus AM4 boards have slightly different socket placement too. My Strix -F actually has a lower Z height than the other two boards crazily enough..

I can tell with my Le Grand Macho RT.. nice and tight on 2 out of 3 boards..

Yeah the Asus placement is necessary for ITX. Otherwise coolers start sticking out the top of small cases which is a huge problem. Unfortunately only ASRock and Asus do it, which is problematic because the two of them are currently dead last in my books......

The only downside I can think of is that the D15 is all but guaranteed to not fit, but that's a pretty tiny tradeoff compared to the throng of coolers that Gigabyte and MSI socket placement make incompatible.

It's been an hour, still holding steady at 67 watts......still got a power Perfcap but now it makes sense as it's actually bouncing off the 70W limit. Thermal equilibrium is about 77C edge. On the Aorus it usually throws in the towel around 10 minutes in and goes directly to 30W.

a2000 stress test.png
 
This is some extreme SFF
 
judging by the stage of debugging you're at rn, you probably havent gotten to ocing the A2000 yet, ... right?

I might not do any at all, I know w1zz had good luck with it but this is just to supplant the Vega 8. It's plenty for the TV duties I ask of it

Would like to get the cooling sorted out since the L9x65 is worse than BS3 and clearly affects both CPU and GPU thermals, but at this point it may be prudent to take my own advice - if it ain't broke don't effing fix it

This is some extreme SFF

It's not extreme SFF without machining a custom single slot cooler for the A2000 and cramming it all into a case at most 75% of this one's size......... :D but with an eye to practicality, 4.6L is as small as I go.
 
I might not do any at all, I know w1zz had good luck with it but this is just to supplant the Vega 8. It's plenty for the TV duties I ask of it
[ ... ]
fair
all i know is, i've recently gotten an A2000 and my OC always fails when theres a load that, for some reason makes it go to like 2,200 mhz or something
like, i've gotten metastable OCs that'd always fail the timespy demo at the exact same place (when the spy opens the time lens and the secondary scene is popping up), but after the demo crashes it'd complete the actual scoring runs w/o any issues

and yeah, w1zz's heaven runs don't really count that's all but guaranteed to be one of those metastable ones that'll crash & burn on basically, anything else (it might not even pass fucking ubm, lmfao) - realistically your core will cap at something like +250-275mhz depending on silicon lottery, w/o shuntmodding it or something
memory doing 2,000 mhz effective (as shown on gpu-z) should be easy-peasy but again, might depend on silicon lottery
 
fair
all i know is, i've recently gotten an A2000 and my OC always fails when theres a load that, for some reason makes it go to like 2,200 mhz or something
like, i've gotten metastable OCs that'd always fail the timespy demo at the exact same place (when the spy opens the time lens and the secondary scene is popping up), but after the demo crashes it'd complete the actual scoring runs w/o any issues

and yeah, w1zz's heaven runs don't really count that's all but guaranteed to be one of those metastable ones that'll crash & burn on basically, anything else (it might not even pass fucking ubm, lmfao) - realistically your core will cap at something like +250-275mhz depending on silicon lottery, w/o shuntmodding it or something
memory doing 2,000 mhz effective (as shown on gpu-z) should be easy-peasy but again, might depend on silicon lottery

I know the strict 70W cap is supposed to be a safety net, but I am also running on a picoPSU and a Dell 240W brick, so I think I will just stay away :D
 
bummer, would've been fun to share/compare oc results w/ you
but yeah anyways, as long as you don't shuntmod it it won't exceed the 70w limit, and it's really easy to unlock quite a bit of extra potential from the A2000 tbh, it's got quite a bit of oc headroom
 
Ordered one of these for the L9x65 to be able to use the A12x15 again


bummer, would've been fun to share/compare oc results w/ you
but yeah anyways, as long as you don't shuntmod it it won't exceed the 70w limit, and it's really easy to unlock quite a bit of extra potential from the A2000 tbh, it's got quite a bit of oc headroom

This week I'll probably look for a kill-a-watt that I can use to determine just how much power I am drawing from the wall. If we still have headroom, I'll give the OC a shot. Though, most likely it will be an undervolt given the temps that it is already running at.......assuming the A12x15 on the L9x65 does not improve GPU temps

Might give GFE undervolting a try, never used it before
 
I am curious how much juice you are pulling with one of those little guys. So much badassery crammed into a tight space man!. The next step is to form a case around your selected components, :rockout:
 
yeah, i haven't gotten to undervolting yet, all i've done thus far is dragging the core & memory clock sliders to the right
voltage-frequency curve's next on the menu ig, but i've got like, no experience w/ it at all
 
yeah, i haven't gotten to undervolting yet, all i've done thus far is dragging the core & memory clock sliders to the right
voltage-frequency curve's next on the menu ig, but i've got like, no experience w/ it at all

interesting

except that time I ran the OCd 2060 super under water, in the past 4 gens of Nvidia GPUs I have only ever done V-F stuff, never sliders

it's pretty easy, just lop off the top into a flat line starting at the volt point you want
 
just clicked around my afterburner. apparently i can't seem to unlock the voltage at all. feelsquadroman.
 
Another reminder that the road to SFF is fraught with unforeseen peril......

View attachment 310521View attachment 310522

In the middle of a game and the RTX A2000 suddenly decides to limit itself below 30W of total board power and gets stuck there. As expected, massive lag and panic ensues. Never happened before when running Timespy or games.

The card just refused to exceed 30W. Core and mem both constantly spiking to 3D clocks then immediately dropping back again. %TDP Power Consumption in GPU-Z would start normal then as soon as 3D load began it would blank out and display nothing. Constant Pwr perfcap. Nothing showed as amiss in software for the A2000.
  • Tried clean installing a bunch of different drivers
  • DDU'd
  • Removed leftover AMD driver garbage (thanks ddu NOT)
  • Reflashed BIOS
  • fTPM off and on
  • Updated to 22H2
  • HAGS off and on
  • PCIe link speed
In the end it was some combination of the following that got the card back to normal:
  • Cleaned out the card
  • Cleaned out x16 slot
  • Reseated the card
  • Made a 76W PBO profile for the 5700G
  • Cleared CMOS
  • Unplugged and discharged mobo power
Since it is a slot-powered card after all, stands to reason gunk in the PCIe slot could have had something to do with it.
I've ran into situations where a driver can crash out (happened a lot with my memory overclocking testing and with GPU undervolt testing) and the GPU would get stuck in weird profiles, with afterburner having no effect on clock speeds.
Usually a full shutdown and PSU disconnect was enough to get things working properly here, plausible you had the same issue (even if the driver crash was from a different cause)
 
yeah im aware of that like, i've already got an OC on my A2000 going, but it's only the classic, easy one where you just drag your core/memory freq offset and hope it doesn't crater during load

i just can't seem to be able to unlock the voltages; im on the recentmost drivers & afterburner (wait jk i didnt update to 537.13 yet, it's still 536.99)
 
Back
Top